Molecular Basis for Peptidoglycan Recognition by a Bactericidal Lectin

Molecular basis for peptidoglycan recognition by a bactericidal lectin

Rebecca E. Lehotzkya,1, Carrie L. Partchb,1, Sohini Mukherjeea, Heather L. Casha, William E. Goldmanc, Kevin H. Gardnerb,2, and Lora V. Hooperd,a,e,2

dThe Howard Hughes Medical Institute and Departments of aImmunology, eMicrobiology, and bBiochemistry, University of Texas Southwestern Medical Center, Dallas, TX 75390; and cDepartment of Microbiology and Immunology, University of North Carolina, Chapel Hill, NC 27599

Edited* by Stuart A. Kornfeld, Washington University School of Medicine, St. Louis, MO, and approved February 11, 2010 (received for review August 19, 2009)

RegIII proteins are secreted C-type lectins that kill Gram-positive The bactericidal activities of RegIIIγ and HIP/PAP depend on bacteria and play a vital role in antimicrobial protection of the binding to cell wall peptidoglycan, a polymer of alternating mammalian gut. RegIII proteins bind their bacterial targets via N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid interactions with cell wall peptidoglycan but lack the canonical (MurNAc) cross-linked by short peptides. This finding identified sequences that support calcium-dependent carbohydrate binding RegIII lectins as a distinct class of peptidoglycan binding proteins in other C-type lectins. Here, we use NMR spectroscopy to deter- (1). However, given the lack of canonical carbohydrate binding mine the molecular basis for peptidoglycan recognition by HIP/ motifs in RegIII proteins, the molecular basis for lectin-mediated PAP, a human RegIII lectin. We show that HIP/PAP recognizes the peptidoglycan recognition remains unknown. Furthermore, peptidoglycan carbohydrate backbone in a calcium-independent RegIII lectins are secreted into the intestinal lumen where there manner via a conserved “EPN” motif that is critical for bacterial kill- are abundant soluble peptidoglycan fragments derived from the ing. While EPN sequences govern calcium-dependent carbohydrate resident microbiota through shedding or enzymatic degradation recognition in other C-type lectins, the unusual location and of the bacterial cell wall. Thus, it remains unclear how RegIII calcium-independent functionality of the HIP/PAP EPN motif lectins selectively bind to bacterial surfaces without competitive suggest that this sequence is a versatile functional module that inhibition by soluble peptidoglycan fragments in the lumenal

can support both calcium-dependent and calcium-independent environment. BIOCHEMISTRY carbohydrate binding. Further, we show HIP/PAP binding affinity In this study, we use solution NMR spectroscopy to understand for carbohydrate ligands depends on carbohydrate chain length, the molecular basis of peptidoglycan recognition by HIP/PAP. “ ” supporting a binding model in which HIP/PAP molecules “bind First, we identify a HIP/PAP EPN motif that is essential for and jump” along the extended polysaccharide chains of peptido- recognition of the peptidoglycan carbohydrate backbone and is required for efficient bacterial killing. Whereas EPN tripeptides glycan, reducing dissociation rates and increasing binding affinity. 2þ We propose that dynamic recognition of highly clustered carbohy- are involved in Ca -dependent recognition of carbohydrates by other C-type lectins (5, 6), the HIP/PAP EPN sequence differs in drate epitopes in native peptidoglycan is an essential mechanism 2þ governing high-affinity interactions between HIP/PAP and the that it supports Ca -independent saccharide recognition and is bacterial cell wall. found in an unusual location in the protein. Second, we show that HIP/PAP binding affinity to soluble peptidoglycan analogs antimicrobial protein ∣ C-type lectin ∣ intestine ∣ nuclear magnetic depends on carbohydrate chain length, pointing to a model in resonance ∣ bacterial cell wall which multivalent presentation of carbohydrate epitopes in native peptidoglycan governs high-affinity interactions between HIP/PAP and the bacterial cell surface. he intestinal epithelium is the major interface with the indig- Tenous microbiota and a primary portal of entry for bacterial Results pathogens. To defend against continual microbial challenges, RegIII Family Members Lack Canonical Carbohydrate Binding Motifs. intestinal epithelial cells produce a diverse repertoire of secreted The three-dimensional structure of HIP/PAP exhibits a C-type protein antibiotics that protect against enteric infections and limit lectin fold with its characteristic “long loop” structure (7). This opportunistic invasion by symbiotic bacteria. loop exhibits two distinct subdomains, which we have designated We previously identified lectins as a distinct class of secreted “Loop 1” (residues 107–121) and “Loop 2” (residues 130–145) antibacterial proteins made in the mammalian intestinal epithe- (Fig. 1A). In other C-type lectins, such as mannose binding pro- lium (1). Members of the C-type lectin family share a common tein (MBP) and DC-SIGN, a tripeptide motif in Loop 2 (EPN or protein fold, the C-type lectin-like domain (CTLD), that is QPD) is critical for Ca2þ-dependent binding to saccharides, defined by a set of conserved sequences (2). Many C-type lectin 2þ 2þ interacting with both Ca and carbohydrate to coordinate the family members bind carbohydrate ligands in a calcium (Ca )- sugar 3-OH (6) (Fig. 1B). A second motif (ND) located in the dependent manner through a distinct set of conserved residues β4 strand also makes contacts with both Ca2þ and the 4-OH that govern both Ca2þ-dependence and carbohydrate ligand specificity (2). The Reg (regenerating) family constitutes a unique group of mammalian CTLD-containing proteins that consist Author contributions: R.E.L., C.L.P., S.M., H.L.C., K.H.G., and L.V.H. designed research; R.E.L., ∼16 C.L.P., S.M., and H.L.C. performed research; W.E.G. contributed new reagents/analytic solely of a -kD CTLD and N-terminal secretion signal. tools; R.E.L., C.L.P., S.M., and H.L.C. analyzed data; and R.E.L., C.L.P., K.H.G., and L.V.H. Despite having a canonical C-type lectin fold, Reg proteins lack wrote the paper. 2þ conserved sequences that support Ca -dependent carbohydrate The authors declare no conflict of interest. binding in other C-type lectins (2). Several RegIII proteins are *This Direct Submission article had a prearranged editor. γ highly expressed in the small intestine, including mouse RegIII Freely available online through the PNAS open access option. and human HIP/PAP (hepatointestinal pancreatic/pancreatitis 1R.E.L. and C.L.P. contributed equally to this work. associated protein) (1, 3). We have shown that RegIIIγ and 2To whom correspondence may be addressed. E-mail: [email protected] or HIP/PAP are directly bactericidal for Gram-positive bacteria [email protected]. at low micromolar concentrations (1, 4), revealing a unique bio- This article contains supporting information online at www.pnas.org/cgi/content/full/ logical function for mammalian lectins. 0909449107/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.0909449107 PNAS Early Edition ∣ 1of6 Downloaded by guest on September 30, 2021 0.06 A Loop 1 Loop 2 A Loop 1 Loop 2 E118 E114 β4 0.04 (ppm) TOT

∆δ 0.02

0.00 27 77 127 177 C Residue B N 112 saccharide 122 117 B Loop 2 β4 specificity 0 0.1 γ E118 124 0.3 114 119 0.6 E114 N (ppm)

15 0.9 δ 1.3 N116 N 2 126 mM sPGN 116 121 7.7 7.5 8.0 7.8 8.3 8.1 Fig. 1. HIP/PAP lacks canonical C-type lectin-carbohydrate binding motifs. 1H (ppm) (A) NMR structure of HIP/PAP (PDB code 2GO0.pdb) (22), with Loops 1 and 2 of the long loop region in Red and the β4 strand in Blue.(B) Alignment of C the long loop region of RegIII family members with other C-type lectin family members. MBP-C and DC-SIGN harbor Loop 2 EPN motifs that govern sugar ligand binding and confer selectivity for mannose (Man) and GlcNAc (5). Asia- loglycoprotein receptor (ASGP-R) and aggrecan contain Loop 2 QPD motifs that confer selectivity for Gal and GalNAc. Despite their selective binding to GlcNAc and Man polysaccharides (1), RegIII lectins lack the Loop 2 EPN motif. apo DE1 mM CaCl of the carbohydrate (5, 6). Sugar binding and specificity is dic- 2

tated by positioning of the hydrogen-bond donors and acceptors 124 26 N116 Cγ in the residues flanking the conserved proline, and is determined 115 by the orientations of the sugar 3- and 4-OH groups. Selective 30 C (ppm) N (ppm) 125 β 13

15 C binding is conferred by the Loop 2 tripeptide motif with the E114 34 EPN motif binding GlcNAc or mannose (which have equatorial 116 3- and 4-OH groups), and the QPD motif binding GalNAc or 7.6 7.4 8.8 8.6 8.8 1H (ppm) 1H (ppm) galactose (which have axial 4-OH groups) (5) (Fig. 1B). While RegIIIγ and HIP/PAP selectively recognize GlcNAc or mannose Fig. 2. Peptidoglycan induces HIP/PAP Loop 1 conformational changes that polysaccharides, they do so in a Ca2þ-independent manner (1, 8). encompass a conserved EPN motif. (A) Quantification of HIP/PAP chemical 15 1 Furthermore, both proteins lack the Loop 2 EPN and the β4ND shift changes from the N∕ H HSQC in the presence of 1.3 mM solubilized motifs (Fig. 1B), suggesting a distinctive mode of carbohydrate Staphylococcus aureus peptidoglycan (sPGN). The Red Dotted Line indicates chemical shift changes greater than the mean þ2σ (0.029 ppm). (B) 15N∕1H recognition. 15 HSQC spectra of 100 μM N-HIP∕PAP titrated with increasing concentrations of sPGN. Nδ2, side-chain amide. (C) E114 is part of a HIP/PAP Loop 1 EPN motif Peptidoglycan Induces Conformational Changes in a Loop 1 EPN Motif. that is conserved in RegIIIγ but is lacking in Ca2þ-dependent C-type lectins. We used solution nuclear magnetic resonance (NMR) spectro- (D) 15N∕1H-HSQC spectra of 100 μM 15N-HIP∕PAP in the absence and presence scopy as an unbiased approach for identifying HIP/PAP residues of 1 mM CaCl2. The lack of chemical shift perturbations indicates no isomer- involved in peptidoglycan binding. Chemical shift perturbations ization of the E-P peptide bond in the presence of Ca2þ.(E) C(CO)NH TOCSY 13 15 of backbone amide resonances, observed in 15N∕1H HSQC spec- strip of C∕ N-HIP/PAP P115. The separation between the Cγ and Cβ peaks is tra, are sensitive indicators of changes in chemical environment at 5.64 ppm, which is within the expected range for a trans peptide bond (10). specific amino acid residues. We titrated solubilized peptidogly- Dashed Circles indicate the cross-peak locations predicted for a cis peptide bond. can (sPGN) into 15N-labeled HIP/PAP and monitored chemical shift and linewidth changes in 15N∕1H HSQC spectra. As we had consistent with this group serving as a critical hydrogen-bond previously assigned the backbone chemical shifts of 98% of the ’ HIP/PAP 15N∕1H resonances (4), we were able to map these donor to the sugar 3 -OH group in other lectins (7). Because changes to specific sites within the protein. proline residues lack backbone amides, P115 is not detectable in the 15N∕1H HSQC spectrum. During titrations with sPGN, we did not observe significant 2þ chemical shift perturbations in Loop 2, where most C-type lectins In C-type lectins that require Ca to bind carbohydrate, the glutamic acid and asparagine residues of EPN function to coor- have been shown to bind their carbohydrate ligands. However, we 2þ detected small but significant (Δδ > 0.04 ppm) dose-dependent dinate Ca (6). Notably, the peptide bond between the E and P residues of the EPN motif of MBP undergoes a reversible switch chemical shift changes in five backbone amides within Loop 1: 2þ G112, T113, E114, W117, and E118 (Fig. 2A, B). Loop 1 has from a trans to cis conformer in the presence of Ca , orienting the protein backbone for Ca2þ coordination and carbohydrate not previously been ascribed any function in ligand recognition 2þ by C-type lectins. binding (9). Because RegIII lectins do not require Ca to bind Inspection of the HIP/PAP primary sequence revealed that peptidoglycan, chitin, or mannan (1, 8), we asked whether the 2þ HIP/PAP E114 is part of a Loop 1 EPN sequence that is con- HIP/PAP E-P bond also isomerizes in the presence of Ca . 15 1 served in HIP/PAP and RegIIIγ but is absent from Loop 1 of The HIP/PAP N∕ H HSQC spectrum revealed no perturbation C-type lectins that recognize mannose and GlcNAc in a Ca2þ-de- of either the E114 or N116 cross-peaks upon Ca2þ addition, pendent manner (Fig. 2C). Cross-peaks from the N116 side-chain indicating that the E-P peptide bond does not isomerize in the amide also shifted significantly with titration of sPGN (Fig. 2B), presence of Ca2þ (Fig. 2D). To determine the E-P peptide bond

2of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0909449107 Lehotzky et al. Downloaded by guest on September 30, 2021 configuration, we performed C(CO)NH TOCSY on 13C∕15N-la- A 0.8 wild-type B 100 13 13 E114Q K =0.6 mM beled HIP/PAP to obtain the P115 Cβ and Cγ chemical shifts. d 0.6 E118Q The difference between these chemical shifts reports on the 10 K =0.4 mM configuration of the peptide bond preceding the proline and 0.4 d was measured at 5.64 ppm in the absence of Ca2þ, indicative

∆δ (ppm) 1 of the trans conformer (10) (Fig. 2E). Collectively, these findings 0.2 Kd=2.1 mM

suggested that the EPN sequence might be important for ligand % CFU Remaining 2þ recognition in the RegIII lectins, but with a unique Ca -inde- 0.0 0.1 pendent functionality. 0123 012345 µ We fitted the Loop 1 chemical shift perturbations in the pre- mM sPGN M Protein sence of sPGN to extract a binding affinity of 400 μM (Table 1). Fig. 3. HIP/PAP E114 is essential for peptidoglycan binding and bactericidal 4 This Kd is several orders of magnitude higher (∼10 -fold) than activity. (A) Staphylococcus aureus sPGN titrated into 100 μM 15N-HIP∕PAP the Kd of 26 nM for insoluble native peptidoglycan (1), suggesting Loop 1 mutants. HIP/PAP-E114Q has a reduced affinity for sPGN, as evidenced a molecular basis for selective binding of HIP/PAP to bacterial by a decrease in the dose-dependent chemical cross-peak perturbation of K s cell surfaces despite the presence of soluble peptidoglycan T113 in the E114Q mutant. d were derived from chemical shift changes fragments from the intestinal microbiota. at eight Loop 1 residues and averaged (values SEM are also reported in Table 1). (B) Point mutations in HIP/PAP Loop 1 attenuate bactericidal activity. Listeria monocytogenes was incubated with purified wild-type or mutant HIP/PAP E114 is Essential for Peptidoglycan Binding and Bactericidal HIP/PAP for 2 h at 37 °C and quantified by dilution plating. Assays were done Activity. To test the functional importance of HIP/PAP Loop 1 in triplicate. Mean SEM is plotted. residues in peptidoglycan binding, we introduced conservative Glu (E) to Gln (Q) mutations into E114 and E118, the two Loop included perturbations in both the E114 backbone amide and the 1 residues with the most pronounced chemical shift perturbations N116 side-chain amide peaks (Fig. 4B). This suggests that similar upon sPGN titration (Fig. 3A). Binding of these mutants to sPGN was assessed by monitoring changes in their 15N∕1H HSQC spec- chemical moieties in sPGN and chitopentaose elicit similar conformational changes in the protein, particularly at residues that tra upon sPGN titration. Both mutants retained the same general K chemical shift patterns of wild-type HIP/PAP, indicating that the are components of the Loop 1 EPN motif. The chitopentaose d overall protein structure remained intact (Fig. S1). The E118Q was 5.0 mM, indicating a moderate affinity interaction that is consistent with the millimolar Kds typically observed for C-type lectin

mutation had only a minor effect on sPGN binding affinity BIOCHEMISTRY whereas the E114Q mutation weakened the interaction by interactions with monosaccharides and short oligosaccharides >5-fold (Fig. 3A, Table 1). Thus, E114 plays a key role in peptidoglycan binding, suggesting the importance of the Loop 1 EPN ABLoop 1 Loop 2 0.06 114.25 N116 Nδ2 motif in HIP/PAP peptidoglycan recognition. 122 We next assessed whether the reduced peptidoglycan binding 0 0.04 E114 115.25 of the E114 mutant correlated with HIP/PAP antimicrobial acti- 1

vity. HIP/PAP-E114Q showed a >6-fold reduction in bactericidal (ppm) 124 2 8.0 7.9 N ppm TOT 3 activity against Listeria monocytogenes (Fig. 3B), indicating that 0.02 15 ∆δ 5 E114 E114 is also critical for HIP/PAP bactericidal function. Surpris- saccharide 126 length ingly, we detected a similar reduction in bactericidal activity in 0.00 HIP/PAP-E118Q (Fig. 3B). Thus, while E118 is not essential 27 77 127 177 7.9 7.7 7.5 1H ppm for peptidoglycan binding, it is important for antimicrobial Residue function. Together, these data establish the importance of Loop C 1 residues, including E114, for HIP/PAP bactericidal function. 106 0 HIP/PAP-E114Q 1.3 HIP/PAP E114 is Essential for Binding to the Peptidoglycan Carbohy- 111 drate Moiety. Because the EPN motif is important for carbohy- 4.5 drate binding in other C-type lectins, we postulated that 6.8 residues of the HIP/PAP Loop 1 EPN motif specifically recognize 116 13.5

the peptidoglycan carbohydrate moiety. To test this idea we ex- mM chitopentaose amined HIP/PAP binding to chitooligosaccharides, β1,4GlcNAc 16.5

polysaccharides that mimic the linear structure of the peptidogly- N ppm 15 can carbohydrate backbone (Fig. S2). Titration of chitopentaose 121 15 ðβ1; 4GlcNAcÞ5 into N-labeled HIP/PAP yielded chemical shift changes in several of the same Loop 1 residues (Fig. 4A) and along the same trajectory (Fig. 4B) as for sPGN. These changes 126

Table 1. Binding of PGN analogs to HIP/PAP mutants 131 10 9 8 7 HIP/PAP sPGN Chitopentaose Tri-DAP 1H ppm variant Kd (mM)* Kd (mM) Kd (mM) Fig. 4. Carbohydrate recognition by the HIP/PAP Loop 1 EPN motif depends Wild-type 0.4 ± 0.2 5.0 ± 1.6 6.4 † on saccharide chain length. (A) Quantification of HIP/PAP chemical shift E114Q 2.1 ± 1.6 NB 8.4 changes from the 15N∕1H HSQC recorded in the presence of 20 mM chitopen- E118Q 0.6 ± 0.3 nd nd ‡ taose. The Red Dotted Line indicates chemical shift perturbations greater E114Q/E118Q 2.6 ± 1.0 NB 8.0 than the mean þ2σ (0.024 ppm). (B) Overlay of residue E114 backbone amide NB, no binding detected; nd, not determined and N116 side-chain amide chemical shifts with addition of 20 mM GlcNAc, *Kd s were calculated based on chemical shift changes of eight Loop1 chitobiose, chitotriose, or chitopentaose. (C) Mutation of HIP/PAP E114 residues (SEM). abolishes chitopentaose binding. Superimposed 15N∕1H HSQC spectra of †P<0.0002. 15N-labeled HIP/PAP-E114Q with titrated chitopentaose, showing an absence ‡P<0.0003. of chemical shift perturbations in the mutant protein.

Lehotzky et al. PNAS Early Edition ∣ 3of6 Downloaded by guest on September 30, 2021 13.5 (11, 12). Furthermore, the moderate Kd for a soluble peptidogly- 123 112.8 6.8 can analog is consistent with HIP/PAP’s need to selectively recog- S89 4.9 nize the bacterial cell surface without competitive inhibition by 124 Type II 2.6 mM GMDP soluble peptidoglycan fragments from the intestinal microbiota. 113.8 1.3

N ppm 125 E114 0.68 Addition of GlcNAc monosaccharides up to 20 mM produced 15 Type I 0 no detectable chemical shift perturbations (Fig. 4B). Chitobiose 114.8 1.3 126 δ N116 N 2 3.0 ðβ1; 4-GlcNAcÞ2 also failed to elicit significant chemical shift ðβ1; 4 Þ 6.8 changes, whereas chitotriose -GlcNAc 3 titration up to 7.8 7.5 7.2 8.0 7.8 7.6 13.5 20 mM produced chemical shift perturbations that were too small 1H (ppm) 20.0 mM Chitopentaose to allow accurate determination of binding affinity (Fig. 4B). We were unable to examine binding of chitohexaose or other longer Fig. 5. Carbohydrate is the principal determinant of the HIP/PAP-peptido- 15 ∕1 chitooligosaccharide chains due to their reduced solubility at glycan interaction. N H HSQC spectra overlay highlighting the opposing trajectories of E114 backbone amide and N116 side-chain amide chemical millimolar concentrations. Binding was specific for chitooligosac- ðβ1; 4 Þ shift perturbations in response to titration with chitopentaose or GMDP. charides, as titration of cellopentaose -Glc 5 did not elicit Type I and II chemical shift trajectories are indicated. sPGN induces only Type any significant chemical shift changes. This accords with the I trajectories (Fig. 2B), indicating that carbohydrate is the principal determi- absence of detectable HIP/PAP binding to insoluble cellulose, nant of the HIP/PAP interaction with soluble native peptidoglycan. the source of cellopentaose (1). These results show that HIP/ PAP Loop 1 residues, including E114 and N116, recognize soluble analogs of the peptidoglycan carbohydrate backbone, and designated the trajectory observed for chitopentaose as “Type I” demonstrate that binding affinity is dictated by polysaccharide and that observed for GMDP as “Type II” (Fig. 5). Other chemi- chain length. cally defined peptide-containing ligands, including muramyl We next examined the relationship between HIP/PAP binding dipeptide (MDP), muramyl tripeptide, and tracheal cytotoxin affinity and polysaccharide chain length for saccharide–peptide [a naturally occurring disaccharide-tetrapeptide fragment of chains derived from native peptidoglycan. We fractionated sPGN DAP-type peptidoglycan (13)] also induced Type II perturbations by size exclusion chromatography and isolated two well-resolved in the same Loop 1 residues as GMDP. All chemically defined peaks having average saccharide chain lengths of 5 and 12 sugars, peptide-containing ligands tested bound with low millimolar 15 respectively (Fig. S3A). Titrations into N-labeled HIP/PAP re- Kds (Table S1). These results identify Type II chemical shift tra- vealed that the longer sPGN chains elicited more pronounced jectories as a hallmark of peptide interactions with HIP/PAP. chemical shift perturbations at E114 as compared to the shorter sPGN titration elicits only Type I chemical shift trajectories chains (Fig. S3B). We extracted a Kd of 1.4 mM for binding of the (Fig. 2B), suggesting that carbohydrate interactions predominate 12 saccharide fraction and a Kd of >3 mM for binding of the 5 during HIP/PAP binding to sPGN. saccharide fraction (Fig. S3A and Table S1). By comparison, we Titration of the pure peptide ligand Tri-DAP into the HIP/ calculated an average saccharide chain length of 36 for unfractio- PAP-E114Q mutant yielded a Kd of 8.4 mM, similar to the Kd nated sPGN that binds with a Kd of 0.4 mM. These results of 6.4 mM calculated for the wild-type protein (Table 1). Like- support the idea that the affinity of HIP/PAP binding to native wise, binding to HIP/PAP-E114Q/E118Q produced a Kd of peptidoglycan is governed by saccharide chain length. 8.0 mM. This establishes that HIP/PAP E114 and E118 are dis- We next tested whether E114 is essential for carbohydrate pensable for binding to the peptidoglycan peptide moiety and ligand binding. Chitopentaose titration up to 20 mM into the suggests that the peptide interacts with HIP/PAP at a distinct site HIP/PAP-E114Q mutant yielded no detectable chemical shift and elicits conformational changes at these Glu residues through perturbations (Fig. 4C), indicating a lack of binding. This demon- an indirect mechanism. strates that E114 is critical for binding saccharides with structures To further explore the relative importance of HIP/PAP inter- 15 1 that mimic the peptidoglycan carbohydrate backbone and estab- actions with peptide, we examined N∕ H HSQC spectra lishes the importance of the HIP/PAP EPN motif in carbohydrate obtained in the presence of chemically defined peptide-contain- ligand recognition. ing ligands, comparing them to spectra obtained during titrations of the saccharide ligand chitopentaose (Fig. S6). These compa- Carbohydrate is the Principal Determinant of the HIP/PAP-Peptidogly- risons revealed several peptide-specific chemical shift changes 15 1 can Interaction. The peptidoglycan carbohydrate backbone is in the region of the N∕ H HSQC spectrum corresponding to cross-linked by peptide moieties. To determine whether peptide Asn and Gln side-chain amides (Fig. S6). Low crosspeak inten- might contribute to HIP/PAP recognition of peptidoglycan, we sities in triple resonance datasets precluded us from making examined binding of chemically defined peptidoglycan structural definitive assignments of these side-chain resonances; however, analogs that contain peptide. GMDP (GlcNAc-MurNAc-L-Ala- the peptide-specific alterations in the 15N∕1H HSQC spectrum D-Gln) consists of a disaccharide moiety corresponding to the represent another set of unique hallmarks of peptide interactions repeating saccharide subunit of peptidoglycan but with a trun- with HIP/PAP. Notably, none of these peptide-specific alterations cated peptide (Fig. S2). Because the disaccharide did not produce was evident upon sPGN titration, even at saturating ligand detectable chemical shift changes in the NMR binding assay concentrations (Fig. S6). This supports the idea that peptide in- (Fig. 4B), we reasoned that analysis of the NMR behavior of teractions are not predominant in HIP/PAP binding to sPGN, GMDP would allow us to identify peptide-specific alterations which includes fragments with longer saccharide chains in the HIP/PAP 15N∕1H HSQC spectrum. Surprisingly, GMDP (Fig. S3). Rather, our NMR and mutagenesis studies identify titrations produced a pattern of chemical shift changes that were carbohydrate as the principal determinant of the HIP/PAP- similar to those produced by sPGN, occurring at the same Loop 1 peptidoglycan binding interaction. residues (Fig. 5 and Fig. S4). While GMDP induced chemical shift changes along the same Predicting Peptidoglycan Binding Activity of Other RegIII Family Mem- vector as sPGN and chitopentaose, they were on the opposite bers. We compared the HIP/PAP sequence to that of other RegIII trajectory (Fig. 5 and Fig. S5). This behavior indicates a fast family members to determine whether the presence of a Loop 1 interconversion (microseconds or faster) of Loop 1 between EPN predicts peptidoglycan binding. RegIIIγ and RegIIIβ each two conformations in the apo state, with ligand binding shifting also contain a Loop 1 EPN tripeptide. In contrast, RegIIIα lacks the equilibrium more strongly towards either conformer. To dis- this motif, substituting Gln (Q) for Glu (E) at the first position in tinguish between the opposing chemical shift changes, we have the tripeptide (Fig. 6A). We have previously shown that RegIIIγ

4of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0909449107 Lehotzky et al. Downloaded by guest on September 30, 2021 A Loop 1 to elucidate the exact mechanism by which the HIP/PAP EPN motif supports Ca2þ-independent carbohydrate binding. Never- theless, our findings suggest that the EPN tripeptide is a flexible motif that plays multiple roles in lectin-carbohydrate interactions. pellet supe A limited number of other CTLD-containing proteins also BCkD recognize carbohydrates in a Ca2þ-independent manner via 15 mannan chitin unconj. dextran mechanisms that are distinct from HIP/PAP. Dectin-1 binds to RegIII β kD oligomers of the fungal cell wall saccharide β1,3-glucan. Unlike HIP/PAP, dectin-1 does not have a Loop 1 EPN motif, and bind- 15 15 ing to β-glucan involves interactions with the β3 strands of RegIII α dectin-1 dimers (14). The C-type lectin langerin binds mannose in a Ca2þ-independent manner but does so via a Loop 2 site that Fig. 6. The presence of a Loop 1 EPN motif predicts peptidoglycan binding in is secondary to a canonical Ca2þ-dependent binding site (15). other RegIII family members. (A) Alignments of human and mouse RegIII family members showing a conserved EPN sequence in Loop 1 of RegIIIγ HIP/PAP carries out its bactericidal functions in the intestinal and RegIIIβ but not RegIIIα.(B) RegIIIβ but not RegIIIα binds peptidoglycan. lumen, an environment rich in soluble peptidoglycan fragments Ten μg of recombinant RegIIIβ or RegIIIα was added to 50 μg insoluble Bacillus derived from resident bacteria. Because HIP/PAP bactericidal subtilis peptidoglycan and pelleted. Pellet and supernatant (Supe) fractions activity requires recognition of intact bacteria (1), high-affinity were analyzed by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and recognition of soluble peptidoglycan fragments would be coun- Coomassie blue staining. The lower molecular weight form of RegIIIβ in terproductive by competitively inhibiting cell wall binding. Our the pellet results from cleavage at an N-terminal trypsin site by a peptido- finding that HIP/PAP binds to its native polymeric ligand with β glycan-associated proteolytic activity (1, 4). (C) RegIII selectively binds to a much higher affinity than to soluble fragments provides a GlcNAc and mannose polysaccharides. RegIIIβ was bound to immobilized polysaccharide using a previously described assay (1). After washing, bound molecular explanation for the ability of HIP/PAP to selectively proteins were released by boiling and analyzed by SDS-PAGE and Coomassie bind intact bacteria in the lumenal environment. blue staining. Unconj., unconjugated Sepharose. A low affinity for soluble analogs is a characteristic feature of lectins that bind to linear polysaccharides with repeated epitopes. and HIP/PAP exhibit similar binding specificities, recognizing For example, mucin-binding lectins bind to their native extended 106 peptidoglycan, chitin, and mannan but not dextran (1, 8). We ligands with affinities up to -fold higher than for the corre- assessed the binding properties of RegIIIβ and RegIIIα in sponding monovalent carbohydrates that are bound with milli- BIOCHEMISTRY K pull-down assays with insoluble peptidoglycan. RegIIIβ localized molar ds (12). Thermodynamic studies have elucidated the to the peptidoglycan pellet, indicating binding, while RegIIIα re- mechanism by which mucin-binding lectins achieve selective mained in the supernatant, showing no detectable peptidoglycan high-affinity binding to extended native ligands, showing that binding (Fig. 6B). the lectins "bind and slide" from carbohydrate to carbohydrate We next assayed the carbohydrate binding specificity of epitope, resulting in a gradient of decreasing microaffinity con- RegIIIβ. RegIIIβ was incubated with carbohydrates conjugated stants. This "recapture effect" results in a decreased off-rate and to Sepharose beads, including mannan, chitin, and dextran. an increased apparent binding affinity (16). Thus, a hallmark fea- RegIIIβ bound to mannan and chitin, but not dextran (Fig. 6C), ture of these interactions is a direct relationship between binding exhibiting the same binding specificity as HIP/PAP and RegIIIγ affinity and carbohydrate chain length. Our finding that the (1, 8). Thus, peptidoglycan binding ability correlates with a Loop strength of HIP/PAP binding depends on saccharide chain 1 EPN tripeptide in RegIII family proteins, and binding length strongly suggests that HIP/PAP utilizes a similar binding activities can be predicted by the presence of this motif. mechanism. Therefore, we propose that dynamic recognition of the multivalent carbohydrate backbone of native peptidogly- Discussion can is an essential mechanism governing selective high-affinity Here, we have used solution NMR to gain insight into the interactions between HIP/PAP and the bacterial cell wall. molecular basis for peptidoglycan recognition by HIP/PAP, a bac- The molecular basis of HIP/PAP recognition of peptidoglycan tericidal C-type lectin. Surprisingly, we found that Ca2þ-indepen- differs fundamentally from that of the peptidoglycan recognition dent recognition of the peptidoglycan carbohydrate backbone protein family (PGRPs). Drosophila PGRPs bind with high involves an EPN motif similar to that required for Ca2þ-depen- affinity to soluble PGN fragments shed by bacteria, initiating sig- dent recognition of mannose and GlcNAc-containing saccharides naling cascades that direct expression of antibacterial peptides. by C-type lectins such as MBP. Mutation of the Glu residue Peptidoglycan recognition by PGRPs involves moderate- to (E114) of this tripeptide revealed the functional importance of high-affinity interactions with the peptide moiety, allowing discri- this site in HIP/PAP saccharide binding, peptidoglycan recogni- mination between Gram-positive and Gram-negative bacteria tion, and bacterial killing. Further, the presence of an EPN tri- and the initiation of immune responses specific for either group peptide correlates with peptidoglycan binding activity among (17). HIP/PAP,in contrast, interacts with the carbohydrate moiety mouse and human RegIII lectins. of peptidoglycan, accomplishing high-affinity binding to the 2þ Ca -dependent saccharide recognition by lectins such as bacterial cell wall through multivalent interactions with the highly MBP and DC-SIGN is critically dependent on a Loop 2 EPN clustered presentation of carbohydrate epitopes. motif that coordinates Ca2þ and forms hydrogen bonds with sugar hydroxyls (5, 6). The EPN sequence is an essential deter- Materials and Methods minant of specificity for GlcNAc and mannose, which are related Expression and Purification of Recombinant Proteins. Pro-HIP/PAP was by the orientations of their 3- and 4-OH groups (5). Given that expressed and purified as previously described (4, 8). Point mutations were HIP/PAP also selectively recognizes GlcNAc and mannose poly- introduced using the QuikChange II Site Directed Mutagenesis Kit saccharides (1), we propose that the HIP/PAP Loop 1 EPN motif (Stratagene) and primers harboring the desired mutations. Expression and 2þ purification of mutant HIP/PAP was carried out as for the wild-type protein may govern saccharide specificity. The loss of Ca dependence (4, 8). Expression and purification of RegIIIα and RegIIIβ were carried out as in the Loop 1 EPN-carbohydrate interaction could be due to the described for RegIIIγ (8). absence of additional Ca2þ coordination sites such as the ND dipeptide or the constitutive trans conformer of the EPN proline. Preparation of Solubilized Peptidoglycan (sPGN). Native peptidoglycan was Determination of a high resolution structure of HIP/PAP in isolated from Staphylococcus aureus as previously described (18). Insoluble complex with a peptidoglycan structural analog will be required PGN was resuspended to 10 mg∕mL in 50 mM Tris-HCl pH 8 and sonicated

Lehotzky et al. PNAS Early Edition ∣ 5of6 Downloaded by guest on September 30, 2021 for 30 min. The suspension was heated to 90 °C for 30 min, cooled to 37 °C, for 15N∕13C-labeled samples. HIP/PAP was purified as described (8). NMR data and lysostaphin (Sigma) added to 50 μg∕mL. After overnight incubation at were recorded at 25 °C using a cryoprobe-equipped Varian Inova 600 MHz 37 °C, the preparation was heated to 95 °C for 5 min and centrifuged for spectrometer. Spectra were processed using NMRPipe (20) and analyzed with 10 min at 10,000 g. The soluble fraction was lyophilized, resuspended in NMRView (21). 1 mL milliQ H2O and desalted on a Sephadex G-10 column equilibrated in 25 mM MES pH 5.5, 25 mM NaCl. The sPGN preparation was characterized Antimicrobial Assays. Antimicrobial assays were done using HIP/PAP and HIP/ by size exclusion chromatography (Fig. S3), and total fragment concentra- PAP mutants engineered to lack the inhibitory N-terminal pro-segment (4). tions were estimated by quantifying carbohydrate reducing termini (19). Listeria monocytogenes (strain EGD-e) was exposed to the indicated lectin Average fragment chain lengths were determined from the molar ratio of carbohydrate (determined by absorbance at 218 nm) to reducing termini. concentrations at 37 °C for 2 h, and surviving bacteria were quantified by dilution plating as described (1). Chemically Defined Ligands. Chitooligosaccharides (Seikagaku), cellopentaose (Seikagaku), GMDP (Calbiochem), GlcNAc (Sigma), and Tri-DAP (InvivoGen) ACKNOWLEDGEMENTS. This work was supported by National Institutes of were obtained from commercial sources. Health Grants DK070855 (to L.V.H.), GM081875 (to K.H.G.), and CA130441 (to C.L.P.); the Robert A. Welch Foundation (I-1659 to L.V.H.); and the NMR Spectroscopy. To prepare samples for NMR, Escherichia coli BL21- Burroughs Wellcome Foundation (New Investigators in the Pathogenesis CodonPlus (DE3)-RILP transformed with expression plasmids were grown of Infectious Diseases Award to L.V.H.). S.M. is supported by a Helen Hay 15 15 in M9 minimal media containing 1 g∕Lof NH4Cl for uniformly N labeled Whitney Foundation Postdoctoral Fellowship. C.L.P. was also supported as 13 samples, further substituting unlabeled glucose with 3 g∕Lof C6-glucose a Chilton Fellow.

1. Cash HL, Whitham CV, Behrendt CL, Hooper LV (2006) Symbiotic bacteria direct 12. Dam TK, et al. (2007) Binding studies of α-GalNAc-specific lectins to the α-GalNAc expression of an intestinal bactericidal lectin. Science 313:1126–1130. (Tn-antigen) form of porcine submaxillary mucin and its smaller fragments. J Biol 2. Drickamer K (1999) C-type lectin-like domains. Curr Opin Struct Biol 9:585–590. Chem 282:28256–28263. 3. Christa L, et al. (1996) HIP/PAP is an adhesive protein expressed in hepatocarcinoma, 13. Cookson BT, Tyler AN, Goldman WE (1989) Primary structure of the peptidoglycan- – normal Paneth, and pancreatic cells. Am J Physiol 271:G993–1002. derived tracheal cytotoxin of Bordetella pertussis. Biochemistry 28:1744 1749. 14. Brown J, et al. (2007) Structure of the fungal β-glucan-binding immune receptor 4. Mukherjee S, et al. (2009) Regulation of C-type lectin antimicrobial activity by a dectin-1: Implications for function. Protein Sci 16:1042–1052. flexible N-terminal prosegment. J Biol Chem 284:4881–4888. 15. Chatwell L, Holla A, Kaufer BB, Skerra A (2008) The carbohydrate recognition 5. Drickamer K (1992) Engineering galactose-binding activity into a C-type mannose- domain of Langerin reveals high structural similarity with the one of DC-SIGN but – binding protein. Nature 360:183 186. an additional, calcium-independent sugar-binding site. Mol Immunol 45:1981–1994. 6. Weis WI, Drickamer K, Hendrickson WA (1992) Structure of a C-type mannose-binding 16. Dam TK, Brewer CF (2008) Effects of clustered epitopes in multivalent ligand-receptor protein complexed with an oligosaccharide. Nature 360:127–134. interactions. Biochemistry 47:8470–8476. 7. Zelensky AN, Gready JE (2005) The C-type lectin-like domain superfamily. FEBS J 17. Chang CI, Chelliah Y, Borek D, Mengin-Lecreulx D, Deisenhofer J (2006) Structure of 272:6179–6217. tracheal cytotoxin in complex with a heterodimeric pattern-recognition receptor. 8. Cash HL, Whitham CV, Hooper LV (2006) Refolding, purification, and characterization Science 311:1761–1764. of human and murine RegIII proteins expressed in Escherichia coli. Protein Expres Purif 18. Rosenthal RS, Dziarski R (1994) Isolation of peptidoglycan and soluble peptidoglycan – 48:151–159. fragments. Methods Enzymol 235:253 285. 19. Horn SJ, Eijsink VGH (2004) A reliable reducing end assay for chito-oligosaccharides. 9. Ng KK, Park-Snyder S, Weis WI (1998) Ca2þ-dependent structural changes in C-type Carbohyd Polym 56:35–39. mannose-binding proteins. Biochemistry 37:17965–17976. 20. Delaglio F, et al. (1995) NMRPipe: A multidimensional spectral processing system based 10. Schubert M, Labudde D, Oschkinat H, Schmieder P (2002) A software tool for the on UNIX pipes. J Biomol NMR 6:277–293. 13 prediction of Xaa-Pro peptide bond conformations in proteins based on C chemical 21. Johnson BA (2004) Using NMRView to visualize and analyze the NMR spectra of shift statistics. J Biomol NMR 24:149–154. macromolecules. Methods Mol Biol 278:313–352. 11. Collins BE, Paulson JC (2004) Cell surface biology mediated by low affinity multivalent 22. Ho MR, et al. (2006) Human pancreatitis-associated protein forms fibrillar aggregates protein-glycan interactions. Curr Opin Chem Biol 8:617–625. with a native-like conformation. J Biol Chem 281:33566–33576.

6of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0909449107 Lehotzky et al. Downloaded by guest on September 30, 2021